Right now, when there is a new L1 block finalized, the sequencer "hops on" that block and starts creating L2 blocks that point to the new L1 block.

When there is multiple new L1 blocks, to make sure EVM records every L1 block, the sequencer publishes empty L2 blocks that point to those new L1 blocks, before starting pointing to the head finalized L1 block.

After #1788 we'll have to change this strategy.

For example, if we need to add 3 L1 blocks to BitcoinLightClient contract, we can do so by appending 3 system transactions to the same L2 block!

This is also somehow related to #1718